Drug repurposing and polypharmacology to �ght SARS-CoV-2 through the inhibition of the main protease

Therapeutic options are urgently needed to fight the outbreak of a novel coronavirus (SARS-CoV-2), which causes the COVID-19 disease and is spreading rapidly around the world. Drug repurposing can significantly accelerate the identification of drug candidates suitable for clinical evaluation. Moreover, polypharmacological effects may increase antiviral activity and/or counteract severe complications concurrently affecting COVID-19 patients. Herein, we present the results of a computational drug repurposing campaign in search of potential inhibitors of the main protease of SARS-CoV-2. The screening allowed the selection of 22 promising drugs. Some of them have already entered clinical trials, but the vast majority of the identified compounds are new and have never been considered before. For each repurposed compound, its therapeutic relevance and potential beneficial polypharmacological effects that may arise due to its original therapeutic indication are thoroughly discussed.


Introduction
At present, we are faced with one of the most devastating pandemic crises in human history, the coronavirus disease 2019 (COVID-19), which has affected around 56 million people to date and is responsible for more than 1.38 million deaths (who.int 1 ).Unfortunately, neither a vaccine nor effective therapeutic options are currently available to prevent or cure this disease, which is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (Who.int 1 ).However, Pfizer, in partnership with BioNTech Manufacturing GmbH, has very recently filed their clinically evaluated vaccine to Food and Drug Administration (FDA), for an Emergency Use Authorization (EUA) (Pfizer Inc. 2 ).Nevertheless, it should be noted that the efficacy of vaccines for this disease have yet to be assessed on a large scale, and effective, un-restricted treatments for patients already suffering from the disease are yet to be approved for clinical use by worldwide regulatory agencies.
SARS-CoV-2 is a positive-sense single-stranded RNA virus belonging to the β-genus of the Coronaviridae family (Wan et al., 2020) and is closely related to coronaviruses responsible for the SARS and MERS outbreaks that spread in 2003 and 2012, respectively (Zhu et al., 2020).Although outbreaks deriving from coronavirus infections appear to be recurrent (Zhu et al., 2020), neither specific antiviral drugs nor vaccines are currently available to treat these diseases.Therefore, major efforts are currently focused on developing a vaccine, as well as effective drugs to treat infected patients (Clinicaltrials.gov 3 ).Indeed, several COVID-19 vaccine candidates have entered clinical trials, but none of them has been approved so far.In the meanwhile, many biological targets are now under investigation to develop SARS-CoV-2-specific antiviral compounds, e.g., the RNA-dependent RNA polymerase, the spike protein, and the main protease (Morse et al., 2020).Moreover, many clinical trials based on antivirals developed for other infections are now ongoing, together with experimentation of drugs meant to reduce inflammation and severe respiratory complications (Clinicaltrials.gov 3 ).Clinical investigations include, e.g., the antiviral drugs favipiravir, originally developed in Japan to fight influenza and shown to be effective also against Ebola virus (ClinicalTrials Identifiers: NCT04349241, NCT04303299, NCT04310228), and remdesivir, an Ebola virus RNA-dependent RNA polymerase prodrug inhibitor shown to be effective also against MERS and SARS (currently under evaluation for COVID-19 therapy in more than 50 clinical trials) (Clinicaltrials.gov 4 ).Indeed, remdesivir has very recently been reported to significantly reduce the median recovery time of COVID-19 patients and has been approved by FDA as a therapeutic option in hospital settings (Beigel et al., 2020).Moreover, boceprevir, which is a NS3/4A serine protease inhibitor used in the treatment of chronic Hepatitis C, has also very recently gained the attention of the research community for COVID-19 treatment (Fu et al., 2020).Hydroxychloroquine was also investigated for the treatment of COVID-19 but, clinical trials on this compound were stopped by the World Health Organization in June 2020 (Who.int 5 ).Moreover, hydroxychloroquine did not show significant activity on hospitalized patients with COVID-19, according to recently retrieved clinical data (Group et al., 2020b).The monoclonal antibodies tocilizumab and sarilumab directed against the interleukin-6 receptor (IL-6R) have entered clinical trials (ClinicalTrials Identifiers: NCT04322773, NCT04332913) for the same purpose.Finally, the use of dexamethasone was endorsed by the European Medicines Agency to treat hospitalized patients with COVID-19 (Group et al., 2020a).
The rapid spread of the SARS-CoV-2 pandemic and the lack of specific antiviral drugs suggest that drug repurposing should be the preferred method for rapidly selecting suitable candidates for clinical testing (Senanayake, 2020).Indeed, these compounds are well characterized and already possess optimized pharmacokinetics and safety profiles.In this scenario, computational approaches can provide new opportunities for drug repurposing (March-Vila et al., 2017), allowing the identification of valuable drug candidates to be used either alone or in combination.Moreover, considering the severe life-threatening disease complications, polypharmacological drugs, i.e., single drug molecules that combine antiviral activity with e.g., anti-inflammatory or antithrombotic activity, may be particularly useful (Anighoro et al., 2014).To this end, in this study, we performed an extensive structure-based virtual screening campaign and identified 22 top-candidate approved or experimental drugs as potential inhibitors of the SARS-CoV-2 main protease (M pro ).The DrugBank database (Wishart et al., 2018), including drug metabolites, was docked to the M pro enzyme and the results were post-processed with BEAR (Rastelli et al., 2009), an in-house developed screening tool with a well-documented ability to refine virtual screening results (Rastelli and Pinzi, 2019).The candidate selection process took into special consideration the analysis of drug annotations and reported biological activity information to attain possible favorable polypharmacological effects arising from the original therapeutic indication.Because of their fitting to the SARS-CoV-2 main protease active site, the reported drugs could be readily repurposed to elicit an antiviral response.

Materials and Methods
The recently reported 6LU7 crystal structure of the SARS-CoV-2 main protease (Jin et al., 2020) was first collected from the Protein Data Bank (accessed on March 17 th , 2020) and then prepared for the in silico screening process using the Protein Preparation Wizard (Protein Preparation Wizard, Schrödinger, LLC, New York, NY, 2020).Defaults parameters were used during the protein preparation.Atom types and connectivity issues were fixed, hydrogen atoms were added, and interaction geometries were optimized.The co-crystallized water molecules and the peptide-like PRD_002214 inhibitor (compound N3) were retained during the protein preparation process, while they were removed in the following docking and post-processing phases.
Docking calculations were performed in the active site of the prepared 6LU7 crystal structures using Glide (Glide, Schrödinger, LLC, New York, NY, 2020; Friesner et al., 2004) with the Standard Precision (SP) protocol.Specifically, the receptor grid was first generated on the coordinates of the cocrystallized PRD_002214 ligand, with a box of (10 Å × 10 Å × 10 Å) dimensions (default settings).
Then, the docking protocol was validated by redocking the co-crystallized ligand into its parent crystal structure, with satisfactory results.
Approved drugs, clinical and preclinical candidates, and metabolites were first downloaded from the DrugBank database (www.drugbank.ca,accessed on: March 17 th , 2020), and then prepared for the structure-based calculations using the LigPrep utility (LigPrep, Schrödinger, LLC, New York, NY, 2020).Specifically, ionization states and tautomers at pH values equal to 7±2 were first generated for each ligand in the screening database and then minimized according to the OPLS3e force field.Stereoisomers were also generated for the DrugBank ligands with undefined chiralities.Subsequently, the pre-treated compounds were screened with the validated docking protocol.Finally, the predicted poses were visually inspected, and the first 2000 top-scoring ligand-protein complexes were further processed with the BEAR post-docking tool (Rastelli et al., 2009).
The BEAR protocol consists of three steps based on molecular mechanics (MM) minimization and molecular dynamics cycles, followed by more accurate binding free energy estimation of the refined complex with the MM-PBSA and MM-GBSA methods (Rastelli et al., 2009).Further details on the BEAR post-processing procedure, which was demonstrated to considerably increase the prediction performances in several virtual screening campaigns, are reported by (Rastelli et al., 2009).
A final step of visual inspection of the refined complexes and their comparison with the corresponding poses predicted by Glide aided in the final selection of potential candidates for SARS-CoV-2 main protease inhibition.Analysis of data annotation and literature searches returned top candidates with a potentially beneficial polypharmacology profile.

Results
The SARS-CoV-2 main protease (M pro ), also known as 3C-like protease, is a relevant target for drug repurposing because it plays a crucial role in the maturation of the viral particle (Jin et al., 2020).Indeed, this cysteine protease presents a highly conserved active site in several coronaviruses, such as SARS-CoV and MERS-CoV, and plays a key role in the cleavage of precursor polyproteins translated from viral RNA.
A computational screening workflow (Figure 1) was devised to identify drug candidates able to bind and inhibit the SARS-CoV-2 main protease.To this end, the complete DrugBank database (Release Version 5.1.5,13,227 compounds, accessed on March 17 th , 2020), including drug metabolites, was docked to the crystal structure of the SARS-CoV-2 main protease (PDB ID: 6LU7) (Jin et al., 2020) using Glide (Glide, Schrödinger, LLC, New York, NY, 2020.; Friesner et al., 2004) , and the results were ranked according to the Glide "Docking score" scoring function.
The 2000 top ranking compounds were post-processed with BEAR (Rastelli et al., 2009;Rastelli and Pinzi, 2019).BEAR is an automated procedure that combines structural refinement of docking poses through molecular dynamics, which accounts for protein flexibility, with re-ranking of ligands on the basis of binding free-energies.Binding free energy of the refined complexes are calculated through the molecular mechanics generalized Born surface area (MM-GBSA) and molecular mechanics Poisson Boltzmann surface area (MM-PBSA) scoring functions, which implicitly account for solvation effects (Rastelli et al., 2009).Then, the best candidates were selected according to i) docking and post-docking scores, which provide an estimation of the binding affinity of a ligand to a target, ii) visual inspection of the protein-ligand complexes, and iii) analysis of drug annotations and literature information to repurpose known drugs, clinical candidates or experimental compounds as M pro inhibitors that may benefit from their original therapeutic indications to reduce severe COVID-19 complications, while inhibiting viral particle maturation (beneficial polypharmacology (Anighoro et al., 2014)).
The most promising compounds resulting from the virtual screening are reported in Table 1, along with their DrugBank IDs, the predicted docking and binding free energy scores, the rank position in the screening outcome, the therapeutic targets and original indications, and the PDB IDs of the crystal structure of each drug in complex with its original target, when available.The 2D chemical structures of the candidates are reported in Figure S1 (see Supplementary Material).A detailed discussion of the selected drug candidates divided into candidates for drug repurposing, candidates for drug repurposing with beneficial polypharmacology, and candidates for drug repurposing based on active drug metabolites, is reported in the following paragraphs.

Candidates for drug repurposing
The drugs reported herein (either approved or investigational) are excellent examples of potential candidates for repurposing as SARS-CoV-2 M pro inhibitors.Indeed, all compounds are well accommodated in the SARS-CoV-2 M pro catalytic site.Moreover, as evidence of the reliability of our results, several of the proposed candidates are viral protease inhibitors, which in some cases are already under study as COVID-19 therapeutic agents (see below).

MMI-175 (DB02378
) is an experimental drug that inhibits β-secretase (BACE-1) (Ghosh et al., 2005), one of the two aspartic proteases responsible for the generation of amyloid-β peptides in neurons.As such, drugs blocking this enzyme may aid in slowing down Alzheimer's disease progression.
According to the predicted docking pose (Figure 2, panel (A)) and binding affinities, this compound is expected to efficiently bind to the SARS-CoV-2 M pro enzyme.The ability of this compound to cross the blood-brain barrier would be of great interest for COVID-19 treatment, as previous studies have reported the presence of coronavirus particles in the CNS and their potential association with neurologic manifestations in patients (Baig et al., 2020).
Compounds I2 (DB04692), N1 (DB04710) and N3 (PRD_002214, DB04595), three experimental inhibitors of the SARS-CoV M pro reported in 2005, may also be valuable candidate inhibitors of the SARS-CoV-2 M pro .Compounds I2, N1, and N3 were co-crystallized in complex with the SARS-CoV M pro (PDB IDs: 2D2D, 1WOF, and 2AMQ, respectively).Notably, compound N3 has been very recently confirmed to bind and inhibit the SARS-CoV-2 M pro , and the crystal structure of its complex was the starting point of our virtual screening (PDB ID: 6LU7) (Jin et al., 2020).
Compound WRR-183 (DB08732) is an α,β-epoxyketone that irreversibly inhibits the SARS-CoV M pro (PDB code 2OP9) (Goetz et al., 2007).According to our docking results, this compound was predicted to bind to the SARS-CoV-2 M pro tightly and with a similar binding mode, where the epoxide is in close proximity to the Cys residue (Figure 2, panel (B)).Moreover, WRR-183 and especially its C-2 (R) epoxide isomer WRR-182 are highly active against the SARS-CoV spike-mediated entry (Zhou et al., 2011).Therefore, they show the potential to block SARS-CoV and, hopefully, SARS-CoV-2 at two different steps of the replication cycle, i.e., viral entry and particle assembly (Zhou et al., 2011).
Difelikefalin (formerly known as CR-845, DB11938) is a highly selective agonist of the κ-opioid receptor (Keppel Hesselink, 2017).This analgesic opioid peptide acts peripherally and is under investigation for the treatment of acute and post-operative pain and, more recently, chronic pruritus (Keppel Hesselink, 2017).Difelikefalin is currently in two Phase II clinical trials for the treatment of pruritus in atopic dermatitis and biliary cholangitis (Clinical Trials Identifiers: NCT04018027 and NCT03995212).The peripheral analgesic activity of the compound, together with its potential SARS-CoV-2 M pro activity, may prove beneficial to COVID-19 patients experiencing peripheral neurologic symptoms and pain.

Candidates for drug repurposing with beneficial polypharmacology
Polypharmacological ligands are extremely interesting in drug repurposing, because they offer the potential for higher efficacy and a combination of synergistic effects (Anighoro et al., 2014).Therefore, for each top-ranking compound, we carefully investigated whether a possible beneficial polypharmacological effect may arise owing to the reported biological activities and original therapeutic indications.
Enalkiren (DB03395) belongs to the class of direct renin inhibitors.By mimicking the transition state of angiotensin, enalkiren is able to block the first step of the renin-angiotensin system (Glassman et al., 1990).It has been recently reported that SARS-CoV-2 binds to the widespread angiotensin-converting enzyme 2 (ACE2) receptor to enter target cells (Bavishi et al., 2020) and that levels of serum angiotensin II are considerably increased in COVID-19 patients (Liu et al., 2020b).Therefore, modulation of the renin-angiotensin system by enalkiren, coupled with inhibition of the SARS-CoV-2 M pro , may exhibit beneficial effects to treat COVID-19 patients.According to our docking results, enalkiren is well accommodated within the SARS-CoV-2 M pro binding site (Figure 2, panel (C)).Interestingly, another very recent computational study based on a different workflow also identified enalkiren as a potential candidate to target SARS-CoV-2 M pro , further supporting its selection as a promising candidate for COVID-19 treatment (Liu et al., 2020a).
The calpain inhibitor IV (ZLLYCH 2 F, DB04653) is a covalent inhibitor of the calpain-1 cysteine protease (PDB ID: 1ZCM) that regulates the activity of proteins involved in processes influencing neuronal plasticity, cognition and neurodegeneration (Vosler et al., 2008), and is a potential target for intervention in inflammatory diseases (Cuzzocrea et al., 2000).In our docked structure (Figure 2, panel (D)), the reactive methylene group of the compound is in close proximity to the Cys residue of the SARS-CoV-2 M pro active site.This result indicates that a covalent bond can potentially be formed that would provide specificity and higher affinity over other proteases.Interestingly, in 2004, calpain inhibitor IV was shown to be an active agent against SARS-CoV ( Barnard et al., 2004) and its potential use against SARS-CoV-2 has also very recently been reported in other repositioning studies (Abhithaj et al., 2020).Moreover, other calpain inhibitors have recently been tested for inhibition of the SARS-CoV-2 M pro and among them the calpain inhibitor XII was shown to be the most active with an IC 50 of 0.45 µM (Ma et al., 2020).
Ethylsulfonamide-D-Trp-Gln-p-aminobenzamidine (DB04758) was designed to potently inhibit factor VIIa (FVIIa), which forms a complex with tissue factor (TF) to initiate the extrinsic coagulation cascade (Kadono et al., 2005).Compared with other anti-thrombotic agents, the specific targeting of the extrinsic coagulation provides a decreased risk of bleeding.SARS-CoV-2 infection often has dramatic consequences for the circulatory system (Tang et al., 2020b), with preliminary reports including thrombocytopenia, elevated d-dimer levels, prolonged prothrombin time, and disseminated intravascular coagulation (Han et al., 2020).Our in silico findings suggested that DB04758 could also bind with high affinity to the SARS-CoV-2 M pro .Therefore, this molecule might exhibit a dual activity against two crucial aspects of the SARS-CoV-2 infection.

Z-LY-CMK (DB07571
) is a covalent inhibitor of the ATP-dependent Clp protease proteolytic subunit (ClpP), an enzyme that has recently gained attention as a promising drug target for antibiotic development (Bhandari et al., 2018).If confirmed, the potential of this compound to act as both an antimicrobial agent and a SARS-CoV-2 inhibitor would be particularly useful to treat secondary bacterial infections, potentially affecting COVID-19 patients.Moreover, a structurally similar compound (Z LVG CHN2 in the original paper), previously shown to inhibit the herpes simplex virus cysteine protease, has recently been tested on Vero E6 cells infected with SARS-CoV-2.Its antiviral activity (EC 50 = 0.19 µM) was suggested to be due to the inhibition of the SARS-CoV-2 M pro (Riva et al., 2020).
The thrombin inhibitor BM51.1011 (DB07934) is particularly interesting, as thrombotic complications appear to be an important issue in patients affected by COVID-19 (Han et al., 2020).As the pandemic is spreading, the reported coagulation disorders in COVID-19 patients as in previous SARS and MERS patients should be carefully addressed.The structure of BM51.1011 in complex with the original target (thrombin) is reported in PDB ID: 1UVS.
Delparantag (formerly known as PMX-60056, DB12955) is a top scoring candidate for all the scoring functions.This molecule reverses the anticoagulation effects of heparin by binding to the pentasaccharide group of unfractionated heparin (UFH) and low-molecular-weight heparins (LMWH) (Kuziej et al., 2010).Heparin has gained increasing attention for its ability to prevent blood coagulation in COVID-19 patients affected by severe pneumonia and its concomitant anti-inflammatory effects that result in reduced IL-6 levels (Tang et al., 2020a).However, recent studies have shown that patients with COVID-19 can rapidly develop severe or critical vascular diseases, which may result in venous thromboembolism and bleeding status (Xu et al., 2020).As a close monitoring of venous thromboembolism and bleeding risks is essential in such patients, delparantag may be an effective tool to mitigate bleeding risks while eliciting antiviral activity due to the inhibition of the SARS-CoV-2 M pro .

Candidates for drug repurposing based on active drug metabolites
A unique feature of our repurposing strategy involves the inclusion of drug metabolites among screened compounds.Although they are typically discarded in drug repurposing studies, major metabolites can provide extremely interesting results.For this reason, here we describe some of the top scoring metabolites from our screening.Remarkably, many of these metabolites were shown to bind to the active site of the SARS-Cov-2 M pro with higher affinity than their parent drug.
Ritonavir and lopinavir, two HIV-1 protease inhibitors that are administered in combination, are currently being clinically investigated against COVID-19 (Cao et al., 2020).However, recent research findings showed that lopinavir or its combination with ritonavir are not able to significantly inhibit SARS-CoV-2 M pro up to a concentration of 100 µg/m (Jang et al., 2020).Based on our in silico analyses, neither of these drugs appears to interact favourably with the SARS-CoV-2 M pro , although the N-desmethyl metabolite M7 of ritonavir (DBMET00084) was demonstrated to better interact with the protein owing to its free urea group that establishes favourable hydrogen bonds with the SARS-CoV-2 M pro .
The saquinavir decahydroisoquinoline metabolites M2 (DBMET01550 and DBMET01549) and tbutyl hydroxyl M10 (DBMET01548) were predicted to strongly bind to the SARS-CoV-2 M pro active site according to the docking scores.Interestingly, while saquinavir metabolites are inactive against the HIV-1 protease (Noble and Faulds, 1996), they appear to bind to the SARS-CoV-2 M pro binding site with a more favourable predicted binding energy than saquinavir.This finding may be relevant for the pharmacokinetics and dosing of this antiviral for the treatment of COVID-19.

Discussion
The goal of the present study was to perform a systematic drug repurposing screening of compounds from the DrugBank database for their ability to bind to the SARS-CoV-2 main protease.The performed analyses allowed the selection of 22 drugs or experimental compounds, which can be considered "novel", as they had not yet been included in COVID-19 clinical trials at the time of manuscript submission.Interestingly, many of them could be able to combine a potential antiviral activity with other activities, e.g., antithrombotic or anti-inflammatory activity, due to their activity on the primary target and therapeutic indication.Such "polypharmacological" behaviour, if confirmed, would make the identified candidate drugs extremely attractive to be further evaluated for COVID-19 treatment.Indeed, a single molecule would be able to concurrently exert an antiviral activity and mitigate or abolish COVID-19 comorbidities, the severity of which often leads to patient death.Therefore, in this study the importance of these repurposed molecules is also discussed from a polypharmacologic perspective.Furthermore, we identified a number of drug metabolites that appeared to be stronger binders of the SARS-CoV-2 M pro than the parent drugs (e.g., saquinavir and ritonavir metabolites).To the best of our knowledge, drug repurposing based on drug (major) metabolites represents a novel approach that may offer additional and valuable opportunities to repurpose candidate drugs through modulation of in vivo pharmacokinetics.In addition, the in-depth analysis of currently available literature data, including cell-based results on SARS-CoV-2 infected cells, further substantiate the relevance of the potential M pro inhibitors selected in this work.Given the drastic need for therapeutic options for COVID-19, our results can suggest some key drugs for repurposing.Compound DB04595, which is a SARS-CoV-2 M pro inhibitor, emerged in a previous computational screening and is shown in red.For metabolites, therapeutic targets and indications of their parent drugs are shown.Glide ("Docking Score"), BEAR GB (binding free-energy estimation through the MM-GBSA method) and BEAR PB (binding free-energy estimation through the MM-PBSA method) values were used to estimate the ligand-protein binding affinity (values are reported in kcal/mol).Bibliographic references of the crystal structures are reported in Table S1 of the Supplementary information.

Figure 2
Figure 2 Binding mode of MMI-175 (A), WRR-183 (B), enalkiren (C) and calpain inhibitor IV (D) to the 6LU7 crystal structure.The SARS-CoV-2 M pro binding site residues and the potential repurposed compounds are represented in deepteal and raspberry sticks, respectively.The image was created with PyMol (The PyMOL Molecular Graphics System, Version 2.1.1,Schrödinger, LLC).

TableTable 1 :
Drug candidates selected from the computational repurposing campaign.

Table S1 .
Reported crystal structures of the repurposed candidate compounds.